Phrase-Based Named Entity Transliteration on Myanmar-English Terminology Dictionary
Aye Myat Mon, Khin Mar Soe · 2020
Named entity (NE) transliteration is mainly a phonetically based transcription of names across languages using different writing systems. For the Myanmar language, robust transliteration of named entities is still a challenging task, because of the complex writing system and the lack of data. The Myanmar NE transliteration dictionary has so far developed over 135,255 NE instance pairs of western person, organization and place names. We apply statistical experiments on Phrase-based statistical machine translation (PBSMT) model using 2-Grams, 3-Grams, 4-Grams, 5-Grams and 6-Grams language models in decoding. Different units in the Myanmar script, i.e., characters and syllables are compared. We perform experiments on 1,000 test data set and 1,000 development data set of our proposed dictionary and measure the performance of our system applying bilingual evaluation understudy (BLEU) score. We discuss detailed observations of our experiments in this paper. According to the evaluations, we got the significant results on syllable unit for Myanmar (Myan) to English (Eng) transliteration direction with 89.3% BLEU score and on character unit for English (Eng) to Myanmar (Myan) transliteration direction with 82.0% BLEU score.